Re: "The Kernel is the Problem, Not the Solution"

[Stanislav Sedov <stas@xxxxxxxxxxx>]

On May 14, 2013, at 4:57 AM, Anil Madhavapeddy <anil@xxxxxxxxxx> wrote:
> 
> I wouldn't be surprised at all to find many low-hanging scalability
> problems from testing the Mirage-net stack against thousands (and then
> millions) of connections.  So we do need a few shell scripts to automate
> the VM launch and load generation of the clients and server.  Once the
> issues from these are fixed, I would be very interested in using
> SwiftTest as a regression test (i.e. once 10 million connections work,
> they should stay working).
> 

I agree.  It's a pity there're no opensource easy-to-use load
generators available.  I has actually started working on a load
generator based on the packet description language presented in
your PhD a while ago, but didn't make a lot of progress.  The idea
was to make a traffic generator that will rely on the packet descriptions
as a specification for the protocol (so it is not hardcoded and it will
be easy to extend) and make it easy to specify the actions clients
make via some kind of JIT (or AOT) compiled scripted language.  If
something like a netmap backend is used, I believe it will be possible
to get a very decent performance out of it (most likely much better
than commercial load generators with hardwired protocol implementations
and informally coded state machines).

--
ST4096-RIPE